This workshop introduces physics-informed neural networks (PINNs), a machine-learning approach that can complement or provide an alternative to traditional scientific computing methods such as finite difference, finite volume, and finite element methods. Unlike purely data-driven approaches, PINNs incorporate governing physical equations directly into the training process.
Participants will learn the neural network fundamentals needed to develop PINNs. Through examples involving partial differential equations, participants will see how to build PINNs both from scratch and with an established software framework.
Some programming experience is expected. Examples will use Python, NumPy, PyTorch, and JAX, but prior experience with these specific tools is not required.
